MNRAS 446, 2823–2836 (2015) doi:10.1093/mnras/stu2256

Differences between CO- and calcium triplet-derived velocity dispersions in spiral : evidence for central formation?

Rogemar A. Riffel,1‹ Luis C. Ho,2,3 Rachel Mason,4 Alberto Rodr´ıguez-Ardila,5 Lucimara Martins,6 Rogerio´ Riffel,7 Ruben Diaz,8 Luis Colina,9 Almudena Alonso-Herrero,10 Helene Flohic,11 Omaira Gonzalez Martin,12,13,14 Paulina Lira,15 Richard McDermid,4,16 Cristina Ramos Almeida,12,13 Ricardo Schiavon,2,17 Karun Thanjavur,18 Daniel Ruschel-Dutra,7,12 Claudia Winge8 and Eric Perlman19 Affiliations are listed at the end of the paper Downloaded from

Accepted 2014 October 24. Received 2014 October 24; in original form 2014 May 20

ABSTRACT http://mnras.oxfordjournals.org/ We examine the stellar velocity dispersions (σ ) of a sample of 48 galaxies, 35 of which are spirals, from the Palomar nearby survey. It is known that for ultra-luminous galaxies (ULIRGs) and merger remnants, the σ derived from the near-infrared CO band heads is smaller than that measured from optical lines, while no discrepancy between these measurements is found for early-type galaxies. No such studies are available for spiral galaxies – the subject of this paper. We used cross-dispersed spectroscopic data obtained with the Gemini Near-Infrared Spectrograph, with spectral coverage from 0.85 to 2.5 µm, to obtain σ measurements from the 2.29 µm CO band heads (σ CO) and the 0.85 µm calcium triplet at Universidad de Chile on May 20, 2015 (σ CaT). For the spiral galaxies in the sample, we found that σ CO is smaller than σ CaT, with a mean fractional difference of 14.3 per cent. The best fit to the data is given by σ opt = (46.0 ± 18.1) + (0.85 ± 0.12)σ CO.This‘σ -discrepancy’ may be related to the presence of warm dust, as suggested by a slight correlation between the discrepancy and the infrared luminosity. This is consistent with studies that have found no σ -discrepancy in dust-poor early-type galaxies, and a much larger discrepancy in dusty merger remnants and ULIRGs. That σ CO is lower than σ opt may also indicate the presence of a dynamically cold stellar population component. This would agree with the spatial correspondence between low-σ CO and young/intermediate-age stellar populations that has been observed in spatially resolved spectroscopy of a handful of galaxies. Key words: galaxies: active – galaxies: kinematics and dynamics – galaxies: star formation – infrared: galaxies.

relation a very useful alternative. Cosmological simulations sug- 1 INTRODUCTION gest that the central SMBH evolves together with the host galaxy The empirical relationship between the stellar and plays a fundamental role in its evolution (Di Matteo, Springel (σ ) of the spheroidal component of galaxies and the mass of the & Hernquist 2005; Springel, Di Matteo & Hernquist 2005; Bower (SMBH; M•) at their centre (e.g. Fer- et al. 2006; Nemmen et al. 2007), and this co-evolution may be the rarese & Merrit 2000; Gebhardt et al. 2000) has been extensively mechanism that leads to the M•–σ relation. The intrinsic scatter used to estimate M• in active and inactive galaxies. More direct in the relation (i.e. the range of M• found for a given σ ) therefore determinations of M•, through stellar kinematics within the black contains information about the processes of galaxy and black hole hole’s sphere of influence, or broad emission line measurements, evolution (e.g. Gultekin¨ et al. 2009). Besides the M•–σ relation, are only feasible for a limited number of objects, making the M•–σ stellar velocity dispersion measurements are also relevant for many other astrophysical applications, including the galaxy Fundamental Plane (e.g. Djorgovski & Davis 1987; Dressler et al. 1987;Falcon-´  E-mail: [email protected] Barroso, Peletier & Balcells 2002;Bernardietal.2003; Gebhardt

C 2014 The Authors Published by Oxford University Press on behalf of the Royal Astronomical Society 2824 R. A. Riffel et al. et al. 2003; Valluri, Merrit & Emsellem 2004), the metallicity–σ up to twice the value obtained for the NIR. For the remaining eight relation (e.g. Terlevich et al. 1981; Dressler 1984a; Bender, Burstein non-LIRG mergers, σ CaT is slightly larger than σ CO.Evenlarger & Faber 1993), the V/σ ratio that is an important criterion to discrepancies have been found for single nucleus ultra-luminous determine the dynamical state of early-type galaxies (e.g. Cox et al. infrared galaxies (ULIRGs), for which σ CaT can be three times larger 2006; Cappellari et al. 2007; Naab et al. 2013), etc. For all of these than σ CO (Rothberg et al. 2013). The discrepancies arise because reasons, understanding the factors affecting σ measurements is an σ CO, although conveniently insensitive to dust absorption, probes important issue. a luminous, young stellar disc, whereas σ CaT gives information Measurements of σ in galaxies have traditionally been done about an older stellar population that is more representative of the at wavelengths <1 µm, often using the ‘Mgb’ line at 0.52 µm, galaxy’s overall dynamical mass. CO-based σ measurements imply or the 0.85 µm calcium triplet absorption (e.g. Ho et al. 2009). that ULIRGs cannot be the progenitors of giant elliptical galaxies, Measurements based on stellar absorption lines in the infrared (IR), whereas CaT-based values are consistent with a range of final galaxy on the other hand, can also probe regions that are obscured by dust masses. at optical wavelengths, or different stellar populations from those The above studies have compared optical and CO-based σ revealed in the optical. For these reasons, several recent studies have measurements for early-type (E and S0) galaxies, and for galaxy compared σ values obtained from the fitting of stellar absorptions mergers and ULIRGs. Little information is available, though, for in the optical and in the near-IR (NIR) spectral regions, generally late-type (spiral) objects, so the range of morphological types in using the CO absorption band heads in the H and K bands (Silge & which the large differences between σ CO and σ CaT occur is not yet Downloaded from Gebhardt 2003; Rothberg & Fischer 2010;Vanderbekeetal.2011; known. We aim to rectify this situation by measuring σ CO and σ CaT Kang et al. 2013; Rothberg et al. 2013). in a sample of 48 nearby galaxies: 35 spirals, 7 lenticulars, and 6 Silge & Gebhardt (2003) presented stellar velocity dispersions ellipticals. We do this using slit spectroscopy covering simultane- measured from the 2.29 µm CO band head in a sample composed ously the CaT and 2.29 µm CO spectral regions. In Section 2, we of 18 lenticular (S0) and 7 elliptical galaxies. The resulting sigma describe the sample and the observational data, and in Section 3, values were compared with the literature values derived primarily we discuss the methods used to measure the stellar dispersion in http://mnras.oxfordjournals.org/ from fitting the Ca II lines near 4000 Å, the Mgb lines near 5175 the optical and in the NIR. The results are presented in Section 4, Å, and the 8500 Å Ca II triplet (hereafter CaT). The σ CO values while their implications are discussed in Section 5. The conclusions were found to be up to 30 per cent smaller than σ opt, with a median of the present paper are given in Section 6. difference of 11 per cent1. However, this difference was observed only in the lenticular galaxies; in the elliptical galaxies, the optical and IR measurements were consistent. The authors speculate that 2 OBSERVATIONS AND DATA REDUCTION the difference may be related to the presence and distribution of dust The sample of galaxies used in this work comprises 48 objects in the S0 objects: if the dust is mainly located in the stellar disc, selected from the Palomar spectroscopic survey of nearby galaxies optical σ measurements will be biased towards the less obscured, (Ho, Filippenko & Sargent 1995, 1997), covering a wide range of at Universidad de Chile on May 20, 2015 dynamically hotter bulge component <1 µm. luminosity and nuclear activity type. Some properties of the sample σ Vanderbeke et al. (2011) measured CO in a sample of 19 Fornax are shown in Table 1, while full details of the overall programme, cluster members, composed of roughly equal numbers of elliptical sample, observations, and data reduction are given in Mason et al. and lenticular galaxies. These measurements were compared with (submitted). σ opt from Kuntschner (2000) and found to be consistent, with σfrac = 2 σ −σ Briefly, the spectroscopic data were obtained in queue mode CO opt = 6.4 per cent. The lack of a discernible difference between σopt with the Gemini Near-Infrared Spectrograph (GNIRS) on the σ CO and σ opt in the lenticular galaxies does not agree with the Gemini North telescope. The cross-dispersed mode was used with findings of Silge & Gebhardt (2003). It is, however, consistent with the 32 l mm−1 grating, providing simultaneous spectral coverage a previous study of velocity dispersions in the by from approximately 0.85 to 2.5 µm. We used the 0.3arcsec× Silva, Kuntschner & Lyubenova (2008). 7 arcsec slit, generally oriented along the mean parallactic angle Optical and IR velocity dispersions were also compared by Kang at the time of the observations to minimize differential refraction et al. (2013). This work used the CO absorption bands around 1.6 effects. The data were obtained between 2011 October and 2013 −1 µm, rather than those near 2.3 µm, to derive σ CO for 31 galaxies: May. Due to work done to address an issue with the 32 l mm 19 ellipticals, 9 lenticulars, and 3 spirals. No significant difference grating mount in 2012, the spectral resolution of the spectra ob- was found between σ CO and σ opt (based mainly on the CaT). This tained with the 0.3 arcsec slit after 2012 November differs from that suggests that both sets of lines probe roughly the same , and achieved previously. The spectral resolution before 2012 November provides no evidence for a dynamically cold, obscured population. is 11.6 and 4.4 Å for the K band CO band heads and the CaT re- Rothberg & Fischer (2010) also compared σ CO and σ CaT for a gions, respectively, obtained from the full width at half-maximum set of galaxies. This time, however, they studied 14 galaxy mergers, (FWHM) of the arc lamp calibration spectra. These values corre- accompanied by a control sample of 23 elliptical galaxies. The spond to a resolution in velocity dispersion of ∼65 km s−1 for both measurements for the control sample were taken mostly from the regions. After 2012 November, the resulting spectral resolutions are −1 literature, and σ CO and σ CaT were found to be similar, in agreement 18.2 and 7.4 Å corresponding to ∼100 and ∼110 km s . with the studies above. On the other hand, large differences are The data were reduced using standard procedures.3 To summa- found for the mergers. In particular, for luminous infrared galaxies rize, the raw frames are first cleaned of any electronic striping and (LIRGs; six objects in their sample), the σ derived in the optical is

2 Programmes GN-2011B-Q-111, GN-2012A-Q-23, GN-2012B-Q-80, GN- 1 We use σ CO to denote velocity dispersions derived from the 2.29 µmCO 2012B-Q-112, GN-2013A-Q-16. 3 bands, σ CaT for those based on the CaT lines, and σ opt as a general term to Described at www.gemini.edu/sciops/instruments/gnirs/data-format-and- include all measurements based on lines at λ. reduction/reducing-xd-spectra.

MNRAS 446, 2823–2836 (2015) The σ -discrepancy in spiral galaxies 2825

Table 1. Properties of the galaxies from our sample. Column 1: name of the object; column 2: morphological type from Ho et al. (1997); column 3: nuclear Activity from Ho et al. (1997) – S: Seyfert nucleus, L: LINER, T: transition object, and H: H II nucleus. ‘:’ means that the classification is uncertain. Columns 4–9: stellar kinematic parameters. Column 10: the instrumental configuration used for the observations. Configuration ‘a’ corresponds to observations done before 2012 November and has an instrumental σ of 65 km s−1 for both spectral regions. Configuration ‘b’ corresponds to observations done after 2012 November and has an instrumental σ of 100 and 110 km s−1 for the CO and CaT regions. The uncertainties included for the kinematic parameters are those that PPXF outputs. The dashes in the table are due to the fact that for a few objects, we were not able to get a good fit of the spectrum in one of the spectral regions due to a low S/N ratio or non-detection of the absorption lines.

−1 −1 Object Hubble type Nuclear Activity σ CO(km s ) σ CaT(km s ) h3CO h3CaT h4CO h4CaT Conf.

NGC 205 dE5 pec – 40.7 ± 35.3a 98.3 ± 5.3a 0.00 ± 0.15 0.00 ± 0.05 0.01 ± 0.12 − 0.13 ± 0.03 b NGC 266 SB(rs)ab L1.9 204.4 ± 8.2 248.5 ± 26.3 0.00 ± 0.02 0.04 ± 0.07 0.07 ± 0.03 0.01 ± 0.08 a NGC 315 E+: L1.9 322.9 ± 5.5 362.0 ± 15.0 0.02 ± 0.02 0.03 ± 0.04 − 0.08 ± 0.02 − 0.20 ± 0.03 a NGC 404 SA(s)0-: L2 55.0 ± 18.4a 74.5 ± 22.5 − 0.02 ± 0.13 − 0.02 ± 0.19 0.03 ± 0.20 − 0.03 ± 0.16 a NGC 410 E+: T2 276.4 ± 20.0 – − 0.05 ± 0.05 – − 0.03 ± 0.06 – a NGC 474 (R’)SA(s)0 L2: 164.8 ± 5.4 178.7 ± 7.5 0.03 ± 0.02 0.06 ± 0.04 0.06 ± 0.02 − 0.05 ± 0.04 a NGC 660 SB(s)a pec T2/H 164.7 ± 17.3 – − 0.05 ± 0.06 – 0.12 ± 0.07 – b NGC 1052 E4 L1.9 220.3 ± 4.1 250.6 ± 22.3 − 0.01 ± 0.01 − 0.07 ± 0.05 0.03 ± 0.01 0.04 ± 0.06 a ± ± ± ± ± ±

NGC 1167 SA0- S2 179.0 6.2 160.6 41.7 0.01 0.02 0.03 0.06 0.00 0.03 0.00 0.21 a Downloaded from NGC 1358 SAB(r)0/a S2 182.3 ± 6.8 174.3 ± 21.2 − 0.03 ± 0.03 0.08 ± 0.10 − 0.03 ± 0.03 − 0.09 ± 0.09 a NGC 1961 SAB(rs)c L2 189.9 ± 9.2 249.7 ± 13.9 0.01 ± 0.02 0.03 ± 0.05 0.11 ± 0.03 − 0.07 ± 0.04 b NGC 2273b SB(r)a: S2 105.7 ± 14.4 142.4 ± 8.3 − 0.11 ± 0.06 0.05 ± 0.04 0.03 ± 0.09 0.02 ± 0.05 b NGC 2639 (R)SA(r)a? S1.9 160.4 ± 5.2 – 0.02 ± 0.02 – − 0.04 ± 0.03 – a NGC 2655 SAB(s)0/a S2 145.8 ± 6.5 181.1 ± 7.1 0.02 ± 0.02 0.03 ± 0.03 0.08 ± 0.03 0.01 ± 0.03 b NGC 2768 E6: L2 172.8 ± 3.8 177.3 ± 6.7 0.01 ± 0.01 − 0.05 ± 0.03 0.02 ± 0.02 0.04 ± 0.03 a http://mnras.oxfordjournals.org/ NGC 2832 E+2: L2: 328.7 ± 7.1 254.2 ± 16.9 0.06 ± 0.02 0.02 ± 0.05 − 0.01 ± 0.02 − 0.03 ± 0.05 b NGC 3031 SA(s)ab S1.5 182.5 ± 3.8 149.8 ± 7.1 0.00 ± 0.01 0.04 ± 0.03 0.08 ± 0.01 0.03 ± 0.04 a NGC 3079 SB(s)c spin S2 143.6 ± 4.7 – − 0.01 ± 0.02 – 0.05 ± 0.02 – a NGC 3147 SA(rs)bc S2 229.1 ± 4.2 250.2 ± 11.7 0.00 ± 0.01 0.02 ± 0.03 0.06 ± 0.01 0.03 ± 0.03 b NGC 3169 SA(s)a pec L2 169.4 ± 4.0 191.6 ± 17.5 0.01 ± 0.02 0.13 ± 0.04 0.01 ± 0.02 0.04 ± 0.06 a NGC 3190 SA(s)a pec spin L2 189.2 ± 3.9 202.9 ± 13.6 0.02 ± 0.01 0.00 ± 0.06 0.06 ± 0.01 − 0.07 ± 0.05 a NGC 3607 SA(s)0: L2 213.5 ± 3.6 201.0 ± 13.9 0.01 ± 0.01 0.00 ± 0.07 0.06 ± 0.01 − 0.09 ± 0.06 a NGC 3718 SB(s)a pec L1.9 192.7 ± 5.1 224.0 ± 8.4 − 0.03 ± 0.01 − 0.04 ± 0.03 0.08 ± 0.02 − 0.01 ± 0.03 b NGC 3998 SA(s)ab L1.9 346.9 ± 5.9 331.5 ± 17.7 0.02 ± 0.01 − 0.03 ± 0.04 − 0.02 ± 0.02 0.03 ± 0.04 b NGC 4203 SAB0-: L1.9 176.4 ± 5.9 182.2 ± 8.4 0.00 ± 0.02 − 0.02 ± 0.04 0.03 ± 0.02 − 0.05 ± 0.04 b NGC 4235 SA(s)a spin S1.2 209.6 ± 13.5 156.4 ± 12.3 − 0.10 ± 0.03 0.01 ± 0.05 0.12 ± 0.03 0.06 ± 0.05 b at Universidad de Chile on May 20, 2015 NGC 4258 SAB(s)b S1.9 129.6 ± 3.2 132.4 ± 6.4 − 0.02 ± 0.01 − 0.02 ± 0.04 0.02 ± 0.02 − 0.05 ± 0.04 a NGC 4346 SA0 spin L2: 124.1 ± 7.4 154.4 ± 6.5 0.00 ± 0.02 0.06 ± 0.02 0.07 ± 0.04 0.00 ± 0.03 b NGC 4388b SA(s)b: spin S1.9 103.3 ± 12.4 165.3 ± 20.8 − 0.01 ± 0.05 0.01 ± 0.09 0.01 ± 0.08 0.01 ± 0.09 b NGC 4450 SA(s)ab L1.9 118.8 ± 4.6 136.4 ± 10.9 0.00 ± 0.02 − 0.07 ± 0.05 0.08 ± 0.03 − 0.05 ± 0.07 a NGC 4548 SB(rs)b L2 104.0 ± 8.8 131.6 ± 6.8 0.00 ± 0.03 0.00 ± 0.04 0.04 ± 0.06 0.02 ± 0.04 b NGC 4565 SA(s)b? spin S1.9 151.6 ± 4.0 180.0 ± 5.0 0.03 ± 0.01 − 0.01 ± 0.02 0.02 ± 0.02 − 0.04 ± 0.02 b NGC 4569 SAB(rs)ab T2 106.4 ± 7.6 178.2 ± 7.8 − 0.05 ± 0.03 0.08 ± 0.02 0.03 ± 0.05 0.05 ± 0.03 b NGC 4579 SAB(rs)b S1.9/L1.9 177.5 ± 5.3 174.9 ± 10.8 0.05 ± 0.02 0.03 ± 0.04 0.06 ± 0.02 0.01 ± 0.05 a NGC 4594 SA(s)a spin L2 253.9 ± 3.6 271.4 ± 6.3 0.00 ± 0.01 0.03 ± 0.02 0.06 ± 0.01 − 0.01 ± 0.02 b NGC 4725 SAB(r)ab pec S2: 133.5 ± 3.3 162.4 ± 11.3 − 0.05 ± 0.02 0.27 ± 0.08 − 0.02 ± 0.02 − 0.27 ± 0.07 a NGC 4736 (R)SA(r)ab L2 120.2 ± 6.9 135.4 ± 10.0 0.13 ± 0.04 0.01 ± 0.05 − 0.05 ± 0.04 0.01 ± 0.06 b NGC 4750 (R)SA(rs)ab L1.9 105.9 ± 11.5 155.9 ± 7.3 − 0.01 ± 0.03 − 0.05 ± 0.04 0.08 ± 0.07 − 0.08 ± 0.04 b NGC 5005 SAB(rs)bc L1.9 153.3 ± 6.4 183.1 ± 7.5 − 0.01 ± 0.02 − 0.01 ± 0.02 0.08 ± 0.03 0.06 ± 0.03 b NGC 5033 SA(s)c S1.5 151.0 ± 10.7 147.0 ± 32.7 − 0.12 ± 0.04 0.04 ± 0.16 0.15 ± 0.05 − 0.01 ± 0.22 a NGC 5194 SA(s)bc pec S2 56.3 ± 8.0a 91.8 ± 7.0 − 0.02 ± 0.0 − 0.04 ± 0.05 0.03 ± 0.09 − 0.01 ± 0.05 a NGC 5371 SAB(rs)bc L2 142.8 ± 6.6 159.3 ± 6.8 − 0.01 ± 0.02 0.01 ± 0.03 0.04 ± 0.03 − 0.03 ± 0.04 b NGC 5850 SB(r)b L2 118.0 ± 8.5 179.3 ± 7.2 0.05 ± 0.03 0.02 ± 0.03 0.07 ± 0.05 − 0.05 ± 0.03 b NGC 6500 SAab: L2 177.5 ± 5.6 – 0.00 ± 0.02 – 0.04 ± 0.02 – a NGC 7217 (R)SA(r)ab L2 125.8 ± 5.9 157.6 ± 5.6 − 0.01 ± 0.02 0.04 ± 0.02 0.03 ± 0.03 0.00 ± 0.03 b NGC 7331 SA(s)b T2 137.3 ± 3.2 141.3 ± 6.6 0.03 ± 0.01 − 0.03 ± 0.04 0.01 ± 0.02 − 0.05 ± 0.04 a NGC 7743 (R)SB(s)0+ S2 90.2 ± 4.8 133.4 ± 4.8 0.04 ± 0.02 − 0.13 ± 0.04 0.05 ± 0.04 − 0.20 ± 0.03 a aThe measured σ is smaller than the instrumental σ . b The second CO band was excluded from the fitting due to contamination by the [Ca VIII] 2.322 µm emission. cosmic ray-like features arising from α particles in the camera lens luric absorption lines, and roughly flux-calibrated using the telluric coatings. The files are divided by a flat-field and sky-subtracted standard star spectrum. The individual orders are then combined to using blank sky exposures made between the on-source observa- produce the final spectrum. The extraction aperture used for this tions, and then rectified using pinhole images. Wavelength calibra- work was 1.8 arcsec along the 0.3 arcsec slit, corresponding to a tion is achieved using argon arc spectra, and then a spectrum of few tens to a few hundreds of at the distances of these each order is extracted, divided by a standard star to cancel tel- galaxies.

MNRAS 446, 2823–2836 (2015) 2826 R. A. Riffel et al.

3 METHODS Since the order of the polynomials included is small, they do not introduce any bias in the derived stellar kinematics. In order to obtain the line-of-sight velocity distribution (LOSVD), Although PPXF outputs the errors of the measurements, we also we have used the Penalized Pixel-Fitting (PPXF) method of performed 100 iterations of Monte Carlo simulations where random Cappellari & Emsellem (2004) to fit the CO absorption band heads noise was introduced to the galaxy spectrum, keeping constant the around 2.3 µmintheK band and the Ca II λλ8500, 8544, 8665 S/N ratio and the standard deviation of the 100 measurements for (the CaT) stellar absorptions in the Z band. The best fit of the each galaxy. The errors obtained from the simulations are similar galaxy spectrum is obtained by convolving template stellar spectra to the uncertainties that PPXF outputs. with the corresponding LOSVD, assumed to be well represented by Fig. A1 shows the fits of the galaxy spectra at the region of the CO Gauss–Hermite series. The PPXF method outputs the , band heads (2.25–2.41 µm) for all the galaxies of our sample. The velocity dispersion (σ ), and higher order Gauss–Hermite moments fits reproduce the observed spectra very well for all objects, with (h and h ), as well as the uncertainties for each parameter. The h 3 4 3 the residuals being smaller than three times the standard deviation and h moments measure deviations of the LOSVD from a Gaus- 4 of the continuum emission next to the CO absorptions. Two objects sian curve: the parameter h3 measures asymmetric deviations (e.g. (NGC 2273 and NGC 4388) present strong [Ca VIII] 2.322 µmline wings) and the h quantifies the peakiness of the profile, with h > 0 4 4 emission superimposed on the second 12CO absorption band head, and h < 0 indicating narrower and broader profiles than Gaussian, 4 so this band was excluded from the fitted region. respectively (van der Marel & Franx 1993; Riffel 2010). The fits for the CaT spectral region (8480–8730 Å) are shown in The dominant source of error in the velocity dispersion is usually Downloaded from Fig. A2. They also reproduce the observed spectra well, with the related to the choice of stellar template used to fit the galaxy spec- residuals again generally being smaller than three times the standard trum (e.g. Barth, Ho & Sargent 2002b;Riffeletal.2008; Winge, deviation of the continuum emission next to the CaT. However, for Riffel & Storchi-Bergmann 2009). This can be minimized by using some objects, the residuals are somewhat larger (e.g. NGC 5194, a large stellar template library, instead of the spectrum of a single NGC 5371, NGC 7743, and NGC 7332 – see the bottom panels of star. The PPXF allows the use of several stellar template spectra, Fig. A2) due to the lower S/N ratio of those spectra in this region. http://mnras.oxfordjournals.org/ and varies the weighting of each one to obtain the best match to This results in uncertainties of up to 25 km s−1 in the resulting the observed spectrum, minimizing issues arising from template velocity dispersions. mismatches. The set of templates must include stars that closely For the CO spectral region, we find that σ is smaller than match the fitted galaxy spectrum (e.g. Silge & Gebhardt 2003;Em- CO the spectral resolution of the spectra for three objects (NGC 205, sellem et al. 2004;Riffeletal.2008). For the fitting of the CO NGC 404, and NGC 5194), while for the CaT region, only the absorptions, we used as template spectra those of the Gemini NIR dwarf elliptical NGC 205 has σ smaller than the resolution. These late-type stellar library (Winge et al. 2009), which contains spectra  values are marked by in Table 1 and should be considered highly of 60 stars with spectral types ranging from F7III to M5III, observed uncertain. In particular, for the galaxies NGC 205 and NGC 404, in the K band at a spectral resolution of ∼3.2 Å (FWHM). Ho et al. (2009) measured values lower than our resolution, using As template spectra for the CaT region, we used selected at Universidad de Chile on May 20, 2015 higher resolution data. spectra of stars from the CaT stellar library of Cenarro et al. (2001) with a spectral resolution of 1.5 Å (FWHM). This library con- 4 RESULTS tains spectra of 706 stars with all spectral types and is part of the Medium-resolution Isaac Newton Telescope library (MILES; The major difference and main advantage of the present work, Sanchez-Bl´ azquez´ et al. 2006). In this work, we used only the spec- compared to previous studies, is that we measure both σ CO and tra of stars with signal-to-noise (S/N) ratio larger than 50 in the σ CaT from the same spectra, observed in the same way, with the CaT region in order to avoid the selection of noisy spectra by the same aperture and using the same method, while previous studies PPXF code. The final template library contains 190 spectra, including used their own σ CO measurements and compared them with optical several spectral types. We also tested the NASA InfraRed Telescope σ values from the literature. Resulting measurements of the stellar Facility (IRTF) stellar spectral library (Cushing, Rayner & Vacca velocity dispersion and Gauss–Hermite moments for our sample are 2005; Rayner, Cushing & Vacca 2009) that presents spectra of late- shown in Table 1. µ type stars ranging from 0.8 to 5.5 m at a spectral resolving power In Fig. 1,weshowσ CO versus σ CaT, excluding objects for which of R = λ/λ ∼ 2000, similar to the spectral resolution used in good fits could not be obtained for one of the spectral regions this work. The comparisons of the kinematic parameters obtained (marked by dashes in the Table 1). We find that σ CaT tends to be with the IRTF spectral library are similar to the ones obtained with higher than σ CO, with an average difference of σ CO − σ CaT = the two libraries mentioned above. However, the standard devia- −19.2 ± 5.6 km s−1 (for all morphological types). The error in this tion of the residuals (defined as the difference between the galaxy value was obtained using Monte Carlo simulations and the boot- spectrum and the best-fitting model) and the uncertainties are much strap technique (e.g. Beers, Flynn & Gebhardt 1990) as follows. larger using the IRTF library. This may be related to the S/N ratio of First, 10 000 Monte Carlo iterations were run to determine the ef- the IRTF spectra, and/or their lower spectral resolution compared fect to the uncertainties in σ to the mean sigma difference (σ CO − to the MILES library. We therefore decided to use the Gemini and σ CaT). At each run, random values for σ CO and σ CaT constrained MILES libraries for this work. to be within their measured uncertainties were generated and then Since the spectral resolution of the template stellar spectra the σ CO − σ CaT was calculated. The standard deviation of the for both spectral regions is better than the resolution of our 10 000 simulations of σ CO − σ CaT (σu ) represents the effects observations, we degraded the stellar templates to the same res- of the uncertainties in σ to the σ CO − σ CaT. Then, to evaluate olution as the spectra of the galaxies before running the PPXF to the completeness of the sample and its effect to the mean σ dif- measure the LOSVD. In order to properly fit the continuum emis- ference, we run a bootstrap with 10 000 realizations in which for sion, we allowed PPXF to fit multiplicative Legendre polynomials of each iteration, the σ CO − σ CaT is calculated for a sample selected order 1 to account for any slope introduced by dust/AGN emission. randomly among the galaxies of our sample. The standard deviation

MNRAS 446, 2823–2836 (2015) The σ -discrepancy in spiral galaxies 2827

for observations performed after 2012 November differs between the CaT and CO regions (Section 2). The uncertainty (rms) in the FWHM of the arc lamp lines is smaller than 10 km s−1 for both spectral regions, and in order to ‘correct’ the offset from the one-to- one relation observed in σ in Fig. 1, the FWHM in the CaT region would have to be underestimated by more than 30 km s−1. Thus, uncertainties in the instrumental broadening cannot account for the observed differences in CaT- and CO-based σ values in our sample. As noted in Section 3, template mismatch could also intro- duce systematic uncertainties in the σ measurements (e.g. Silge & Gebhardt 2003;Riffeletal.2008). Usually fewer than 10 template spectra are needed by PPXF to successfully reproduce the CO ab- sorption band heads and the CaT. Although the libraries used here are large and contain spectra of stars with several spectral types, the fits to both spectral regions are dominated by giant and supergiant stars, with a slightly larger contribution from supergiants in the K band. These are the spectral types that are expected to dominate the Downloaded from emission in the NIR, suggesting that PPXF is selecting the appropriate stars with which to fit the spectra. For most galaxies, M-stars are the dominant contributor to the fits in the K band, while K-stars domi- nate in the region of the CaT. The difference between the intrinsic widths of the absorption lines in these stars is ∼10 km s−1,which Figure 1. Comparison between the stellar velocity dispersion obtained from is negligible compared to the overall difference between σ CaT and http://mnras.oxfordjournals.org/ the fitting of the CO band heads (y-axis) and from the fitting of the CaT σ (x-axis). The dashed line shows a one-to-one relation. Filled circles are for CO in the sample. In Appendix B, we show the stellar templates the elliptical and lenticular galaxies of our sample and open circles represent used to fit the spectrum of each galaxy. Additionally, there is no 2 the spiral galaxies. difference in the χ values of the fits between early- and late-type galaxies, suggesting that there is no bias in the choice of template σ − σ   for each type of galaxy. of the simulated CO CaT ( σs ) represents the intrinsic scatter σ −σ  As several galaxies of our sample have a Seyfert nucleus, hot of our sample. Finally, the uncertainty CO CaT is obtained by the 2 2 dust emission may play an important role in the K-band continuum. sum of σ and σ in quadrature, as σ −σ  = (σ + σ ) =  u s CO CaT u s The CO band heads can be ‘diluted’ by this emission, which might − − − (2.6kms 1)2 + (4.6kms 1)2 = 5.6kms 1. This value is sim- introduce an uncertainty in the measurements of σ (e.g. Ivanov

CO at Universidad de Chile on May 20, 2015 ilar to the standard error. If we exclude also the three objects et al. 2000; Riffel, Rodr´ıguez-Ardila & Pastoriza 2006; Kotilainen σ with values smaller than the instrumental broadening, we find et al. 2012). To test whether the PPXF accounts correctly for variations σ − σ =− ± −1 CO CaT 17.7 5.7 km s . of the continuum shape, we have simulated contributions of Planck We also performed a Kolmogorov–Smirnov test to determine if functions with temperatures ranging from 700 to 2000 K to the σ σ CaT and CO differ significantly. We found a statistic significance P continuum emission at the K band. No significant difference in σ CO = 0.2 for our sample, where P ranges from 0 to 1 and small values was found when including dust emission ranging from 1 to 70 per mean that the two data sets are significantly different. The value cent of the total K-band emission. Furthermore, for the objects in of P obtained for our sample indicates that there is a reasonable which we could measure the σ from the H-band CO lines (where the σ σ probability (80 per cent) that CaT and CO present discrepant contribution from AGN-heated dust is smaller; Origlia, Moorwood values. &Oliva1993), it agrees to within 10 per cent with that measured The comparisons of higher order Gauss–Hermite moments h3 from the K band and no systematic difference is found between and h4 obtained from the fitting of the CaT with those of the H-andK-band measurements. Thus, we conclude that the hot dust − = CO band heads show an average differences of h3CO h3CaT emission plays a negligible role in the σ measurements. − ± − = ± 0.01 0.03 and h4 is h4CO h4CaT 0.07 0.03. For both the In galaxies with low values of σ , the fitting of higher or- h3 and h4 parameters, there is no correlation between the values der Gauss–Hermite moments can introduce uncertainties in the σ found from the fitting of the CaT and the ones from the CO band measurements (see the bias parameter of the PPXF program; Cap- heads. The values found for h3 and h4 are small for most of the pellari & Emsellem 2004). To test this, we fitted the spectra of the objects in our sample, indicating that the LOSVD of the stars for galaxies assuming that the LOSVD is well described by a Gaus- the nucleus of these galaxies is reasonably well reproduced by a sian, by fitting only the first two moments. In Fig. 2, we present the Gaussian velocity distribution. comparison of σ values obtained from the fitting of four moments with those obtained from the fitting of two moments. This figure shows that the resulting σ for both the CO and CaT spectral regions 5 ANALYSIS OF THE RESULTS are very similar to those obtained when allowing the PPXF to include To understand the observed difference between σ CO and σ CaT in the the h3 and h4 parameters, suggesting that the inclusion of these pa- galaxies, we performed several tests to identify possible systematic rameters does not affect significantly the σ measurements for our effects in the data or introduced by the fitting procedure. sample. As discussed in Section 3, we were very careful in the estimation Finally, we compare our σ CaT measurements with the optical σ of the instrumental broadening, taking into account the facts that values of Ho et al. (2009) for the same galaxies. Ho et al. (2009) (i) our observations present distinct spectral resolutions depending used two spectral ranges to measure σ : a blue region, from 4200 to on the date of the observations, and (ii) the spectral resolution 5000 Å that includes several Fe lines, and a red region, covering the

MNRAS 446, 2823–2836 (2015) 2828 R. A. Riffel et al.

6 DISCUSSION

We have found a systematic offset between σ CO and σ CaT in the galaxies in our sample, which are primarily spirals. In order to further investigate the σ -discrepancy in late-type galaxies and compare the results with the available studies of other galaxy types, we compiled values for the σ CO and for optical measurements (most of them obtained from the CaT region) from the literature for dis- tinct classes of objects. Table 2 presents the resulting σ opt and σ CO values for elliptical, lenticular and spiral galaxies, and merger rem- nants. Since there is good agreement between the σ values obtained from CaT and those from the fitting of other optical lines (e.g. Barth, Ho & Sargent 2002a; Barth et al. 2002b) and as not all the optical values from the literature were obtained from the CaT region, we will use the nomenclature ‘optical’ velocity dispersion (σ opt)tore- fer to measurements of σ obtained from the CaT or other optical lines.

In order to investigate the relation between σ -discrepancy and Downloaded from morphological classification, we plotted in Fig. 4 σ opt versus σ CO for distinct classes of objects. The largest σ difference is observed for the merger remnants and ULIRGs, for which no correlation is found between σ opt and σ CO and the mean difference is σ CO − σ opt − Figure 2. Comparison of σ values obtained from the fit of two and four =−57.8 ± 13.7 km s 1. Elliptical galaxies follow the one-to-one moments in the LOSVD. relation and no discrepancy between optical and NIR measurements http://mnras.oxfordjournals.org/ −1 is found. The mean difference is σ CO − σ opt =−1.5 ± 4.6 km s . For lenticulars, the mean difference is σ CO − σ opt =−10.8 ± 6.4 km s−1 and it can be seen from the figure that most points are distributed around the one-to-one relation. The best linear fit for lenticular galaxies is given by

σopt = (9.1 ± 13.2) + (0.99 ± 0.08)σCO, (1) shown as a dashed line in the bottom-left panel of Fig. 4. at Universidad de Chile on May 20, 2015 For spiral galaxies, we found a mean difference of σ CO − σ opt = −24.0 ± 4.9 km s−1. Excluding NGC 5194, which presents a sigma value smaller than the spectral resolution of the data, we find the  = . −1 σ same relation, with σCO−σCaT 5 0kms . Most objects have −1 smaller than 200 km s and for these objects σ CO is clearly smaller than σ opt. The best linear equation for spiral galaxies is

σopt = (46.0 ± 18.1) + (0.85 ± 0.12)σCO, (2) where we excluded from the fit the galaxy NGC 5194 (identified in Fig. 4). The main cause of uncertainty in equation (2) is the small range of σ probed by the observations. Further observations are needed to cover the high-σ region (σ 220 km s−1), and higher spectral resolution observations of objects with σ 100 km s−1,in Figure 3. Comparison of σ CaT with those found by Ho et al. (2009)using order to improve the calibration of the equation above. various optical lines. We therefore observe that σ CO and σ opt become both more similar and more correlated in early-type galaxies compared with spirals range 6450–6550 Å where Ca+Fe lines are present. They found and ULIRGs/merger remnants. that both values are in good agreement. In Fig. 3, we show our σ CaT versus the σ values presented in Ho et al. (2009). This comparison 6.1 What is the origin of the sigma discrepancy for late-type shows that 50 per cent of the objects have σ differences smaller than galaxies? 10 per cent and for about 90 per cent of the objects the differences are smaller than 25 per cent, indicating that our measurements are As discussed above and in Section 1, the discrepancy between the in agreement with those from Ho et al. (2009). Differences between stellar velocity dispersion obtained from optical bands and that ob- the measurements may be due to the larger aperture (2 arcsec × tained from the NIR CO absorption band heads is larger for mergers 4 arcsec) used by Ho et al. (2009), as well as differences in the S/N of galaxies and ULIRGs than for early-type galaxies. Rothberg & ratio of the spectra and the exact measurement procedures used. Fischer (2010) found a correlation between the infrared luminosity We therefore conclude that our σ measurements are robust, and (LIR)andσ frac for merger remnants, while no correlation is found that the observed difference between σ CaT and σ CO is not due to for elliptical galaxies. Rothberg et al. (2013) showed that the cor- measurement error. relation found for merger remnants extends to ULIRGs, suggesting

MNRAS 446, 2823–2836 (2015) The σ -discrepancy in spiral galaxies 2829

Table 2. Velocity dispersions compiled from the literature.

−1 −1 −1 −1 Object σ CO(km s ) σ opt(km s ) Ref. Object σ CO(km s ) σ opt(km s )Ref.

Elliptical galaxies Lenticular galaxies NGC 221 71 ± 875± 4 [1] NGC 1023 152 ± 11 205 ± 10 [1] 70 ± 275± 3 [2] 217 ± 5 205 ± 10 [2] 60 ± 869± 2 [4] NGC 1161 274 ± 19 297 ± 17 [1] NGC 315 321 ± 59 310 ± 16 [1] NGC 1375 64 ± 456± 10 [3] 324 ± 59 351 ± 16 [4] NGC 1380 190 ± 17 219 ± 11 [3] NGC 821 195 ± 17 209 ± 10 [1] NGC 1380A 60 ± 955± 9[3] 208 ± 5 209 ± 10 [2] NGC 1381 155 ± 6 153 ± 8[3] 188 ± 17 197 ± 20 [4] NGC 1400 212 ± 12 264 ± 26 [1] NGC 1052 211 ± 20 196 ± 4 [4] NGC 2110 224 ± 49 220 ± 25 [1] NGC 1316 212 ± 20 243 ± 9 [4] NGC 2293 255 ± 44 254 ± 13 [1] NGC 1336 119 ± 896± 5 [3] NGC 2380 164 ± 31 189 ± 9[1] NGC 1339 182 ± 9 158 ± 8 [3] NGC 2681 82 ± 9 111 ± 22 [1] NGC 1344 158 ± 20 166 ± 7 [4] NGC 2768 235 ± 51 198 ± 28 [1] ± ± ± ±

NGC 1351 153 7 157 8 [3] NGC 2787 153 8 210 12 [1] Downloaded from NGC 1373 80 ± 575± 4 [3] 186 ± 3 189 ± 9[2] NGC 1374 207 ± 10 185 ± 9 [3] NGC 3115 272 ± 12 230 ± 11 [2] 181 ± 20 180 ± 8 [4] NGC 3245 206 ± 7 205 ± 10 [2] NGC 1379 130 ± 7 130 ± 7 [3] NGC 3384 151 ± 3 143 ± 7[2] 126 ± 20 127 ± 5 [4] NGC 3998 205 ± 16 297 ± 15 [1] NGC 1399 406 ± 33 375 ± 19 [3] NGC 4150 113 ± 18 132 ± 10 [1] http://mnras.oxfordjournals.org/ 336 ± 20 325 ± 15 [4] NGC 4342 224 ± 5 225 ± 11 [2] NGC 1404 247 ± 22 260 ± 13 [3] NGC 4564 175 ± 7 162 ± 8[2] 204 ± 20 230 ± 10 [4] NGC 4596 139 ± 3 136 ± 6[2] NGC 1407 297 ± 40 283 ± 13 [4] NGC 5195 95 ± 6 175 ± 30 [1] 306 ± 40 285 ± 40 [1] NGC 5866 186 ± 14 139 ± 7[1] NGC 1419 125 ± 5 117 ± 6 [3] NGC 6548 225 ± 47 307 ± 23 [1] 116 ± 20 110 ± 6 [4] NGC 6703 146 ± 42 186 ± 9[1] NGC 1427 155 ± 18 175 ± 9 [3] NGC 7332 148 ± 13 130 ± 10 [1] 174 ± 20 172 ± 8 [4] NGC 7457 63 ± 267± 3[2] NGC 2778 161 ± 4 175 ± 8 [2] NGC 7743 66 ± 12 83 ± 20 [1] NGC 2974 272 ± 19 262 ± 13 [1] IC 1963 49 ± 658± 10 [3] at Universidad de Chile on May 20, 2015 262 ± 19 255 ± 12 [4] ESO 358-G06 55 ± 25 58 ± 11 [3] NGC 3377 144 ± 20 145 ± 7 [1] ESO 358-G59 70 ± 20 54 ± 9[3] 147 ± 4 145 ± 7 [2] SPIRAL GALAXIES 134 ± 20 135 ± 4 [4] NGC 1068 129 ± 3 151 ± 7[2] NGC 3379 235 ± 20 185 ± 2 [4] NGC 3031 157 ± 3 143 ± 7[2] NGC 3607 210 ± 8 229 ± 11 [2] NGC 4258 111 ± 2 115 ± 10 [2] NGC 3608 187 ± 4 182 ± 9 [2] Merger remnants/ULIRGs NGC 4261 286 ± 6 315 ± 15 [2] NGC 1614 133 ± 3 219 ± 3[4] NGC 4291 248 ± 7 242 ± 12 [2] NGC 2418 245 ± 7 282 ± 3[4] NGC 4365 262 ± 20 240 ± 3 [4] NGC 2623 152 ± 4 174 ± 3[4] NGC 4374 290 ± 8 296 ± 14 [2] NGC 2914 179 ± 6 178 ± 2[4] NGC 4459 164 ± 6 167 ± 8 [2] NGC 3256 111 ± 20 239 ± 4[4] NGC 4472 291 ± 20 269 ± 3 [4] NGC 4194 98 ± 25 103 ± 2[4] NGC 4473 186 ± 3 190 ± 9 [2] NGC 5018 243 ± 7 222 ± 2[4] NGC 4486 310 ± 20 361 ± 37 [4] NGC 7252 119 ± 19 160 ± 3[4] 331 ± 11 375 ± 18 [2] Arp 193 143 ± 5 229 ± 4[4] NGC 4649 327 ± 11 385 ± 19 [2] IC 5298 150 ± 28 187 ± 4[4] NGC 4697 172 ± 4 177 ± 8 [2] AM 0612-373 240 ± 9 286 ± 9[4] NGC 4742 104 ± 390± 5 [2] AM 1419-263 262 ± 6 258 ± 3[4] NGC 5128 190 ± 13 145 ± 6 [4] AM 2038-382 207 ± 4 256 ± 5[4] NGC 5812 230 ± 6 248 ± 2 [4] AM 2055-425 137 ± 15 207 ± 7[4] NGC 5845 237 ± 4 234 ± 11 [2] IRAS 02021−2103 143 ± 21 209 ± 8[5] NGC 6251 290 ± 8 290 ± 14 [2] IRAS 05189−2524 131 ± 16 265 ± 7[5] NGC 7052 327 ± 13 266 ± 13 [2] IRAS 12540−5708 117 ± 10 346 ± 9[5] NGC 7619 246 ± 47 296 ± 15 [1] IRAS 17208−0014 223 ± 15 261 ± 5[5] 246 ± 47 296 ± 11 [4] IRAS 23365−3604 143 ± 15 221 ± 6[5] NGC 7743 66 ± 12 83 ± 20 [1] NGC 7626 313 ± 20 265 ± 10 [4] IC 2006 125 ± 10 136 ± 7[3] [1] – Silge & Gebhardt (2003); [2] – Kang et al. (2013); [3] – Vanderbeke et al. (2011); [4] – Rothberg & Fischer (2010); [5] – Rothberg et al. (2013).

MNRAS 446, 2823–2836 (2015) 2830 R. A. Riffel et al. Downloaded from http://mnras.oxfordjournals.org/ at Universidad de Chile on May 20, 2015

Figure 4. Comparison of the σ values in the NIR with the optical values for distinct morphological types. The objects from our sample are shown as filled circles and open circles are for measurements from the literature. The dashed line shows a one-to-one relation and the continuous line is the best linear fit of the data. NGC 205 and NGC 5194, which have σ smaller than the instrumental resolution, are identified in the plots.

that dust might play an important role in the σ CO values for this of σ frac, and that the spiral galaxies fill the gap between early-type kind of object. objects and merger remnants. This smooth trend suggests that dust σCO−σopt Fig. 5 shows a plot of σfrac = versus log LIR for the might play some role in the observed σ -discrepancy, as more warm σopt galaxies with available IR luminosities [from Ho et al. (1997), dust is expected in spiral galaxies than in elliptical galaxies, and Rothberg et al. (2013) and Rothberg & Fischer (2010)]. The values less than in merger remnants and ULIRGs. of LIR were estimated using their infrared fluxes FIR from Ho et al. Following Rothberg & Fischer (2010), we estimate the mass of −14 2 (1997), who defined it as FIR = 1.26 × 10 (2.58S60 + S100)Wm , dust by µ S60 and S100 being the flux densities at 60 and 100 m, respectively.    M S 1.5 Although data with higher angular resolution and wider wavelength dust 2 100 = 0.959S100D 9.96 − 1 , (3) coverage are now available from Herschel and Spitzer telescopes, M S60 we use the IRAS fluxes since they are available for most of our objects, while Herschel and Spitzer data are still not available for where S60 and S100 are the IRAS flux densities at 60 and 100 µmin most of them of them (e.g. Marleau et al. 2006; Sauvage et al. 2010; Jy, respectively, and D is the distance to the galaxy in Mpc (see also Ciesla et al. 2012;DeLoozeetal.2012;Auldetal.2013). Fig. 5 Hildebrand 1983; Thuan & Sauvage 1992). Rothberg & Fischer shows that galaxies with higher LIR also have higher negative values (2010) found that σ frac correlates with Mdust for merger remnants,

MNRAS 446, 2823–2836 (2015) The σ -discrepancy in spiral galaxies 2831

lines is larger than that obtained from optical lines, indicating that the NIR samples an obscuring column larger than the optical spec- tral region (e.g. Moorwood & Oliva 1988; Heisler & De Robertis 1999; Martins et al. 2013a). Assuming a standard Galactic extinction curve (Weingartner & Draine 2001; Draine 2003), the extinction at 0.85 µm (CaT) is about a factor of 5 larger than that at 2.3 µm(CO), and the extinction at 0.52 µm(Mgb) is about a factor of 2 greater again. If the difference between σ CaT and σ CO is due to extinction, we may also expect a difference between σ CaT and σ Mgb.However, Barth et al. (2002a) compared σ CaT and σ Mgb anddidnotfindany systematic difference between them. Secondly, warm (T ∼ 50 K), far-infrared-emitting dust may be associated with star formation in these galaxies. Indeed, the emis- sion of the warm dust is directly correlated with the star formation rate, as it is heated by young stars (Kennicutt 1998; Kennicutt & Evans 2012). Young stars form in disc, and in the absence of major perturba- Downloaded from tions generally remain dynamically cold. Indications of young stars in disc have been found in recent spatially resolved spectroscopy of galaxy nuclei with the Near-IR IFU Spectrograph (NIFS) on Gemini North. Riffel et al. (2010, 2011) and Storchi-Bergmann

σ −σ et al. (2012) carried out stellar population synthesis and found a σ σ σ = CO CaT Figure 5. Fractional difference between CO and CaT ( frac ) σ http://mnras.oxfordjournals.org/ σCaT spatial relation between low- CO and the young/intermediate-age versus LIR. The three objects with σ values smaller than the spectral resolu- stellar population, confirming that σ CO is affected by the presence tion of our data are identified in the plot. of young/intermediate-age stars. The presence of a young stellar population has also been proposed as an explanation of the σ -drop observed in some galaxies (e.g. Emsellem et al. 2001;Marquez´ et al. 2003). However, the comparison of our results with spatially resolved measurements should be taken with caution. While the NIFS data resolve a dynamically cold structure, our single-aperture measurements probe the second moment of the LOSVD, which does not necessarily imply a direct link between the two sets of at Universidad de Chile on May 20, 2015 results. New spatially resolved measurements of both the CaT and CO lines would show whether the effect we are observing here is related to the low-σ regions observed in the works cited above. If the difference between σ CO and σ CaT is due to the presence of a young stellar population in the disc of the galaxies, young stars must contribute more to the CO absorption features than do older stars, and the effect of this population on the σ measured from the CaT and optical lines must be negligible. The connection between the CO and CaT bands and the age of the stellar population is not straightforward, though. For example, the CO bands are relatively strong in both young/intermediate-age stars (AGB/TP-AGB stars) and old ones (M-stars), while hotter, younger stars produce strong CaT bands along with weak CO features (e.g. Maraston 2005;Riffel et al. 2007). On the other hand, previous work has shown that σ CaT is not very sensitive to the stellar population (Barth et al. 2002a, 2002b). Full stellar population synthesis would help to resolve these issues, although different models currently make very different pre- Figure 6. σ versus logM for distinct types of objects. frac dust dictions for the NIR spectral region (e.g. Bruzual & Charlot 2003; Maraston 2005). while no correlation is found for elliptical galaxies. Fig. 6 shows 6.2 Implications for the M•–σ relationship the plot of σ frac versus Mdust for the galaxies studied here. A similar trend to that seen in Fig. 5 is observed in this plot, suggesting As discussed above, the σ -discrepancy observed in our sample of that dust plays a role in the observed σ -discrepancy for spirals and mostly late-type galaxies does not appear in previous studies of merger remnants. early-type galaxies. We can use the measured σ values to evaluate Dust may be relevant to the σ -discrepancy in two ways. First, the impact of the σ -discrepancy on determinations of the mass of as extinction is lower in the K band, the σ CO measurements the central SMBH using the M•–σ relationship. Several studies have could probe a dynamically cold, disc-like component that is more aimed at properly calibrating the M•–σ relation for distinct classes obscured than the dynamically hot bulge stars. Indeed, NIR stud- of objects. Xiao et al. (2011) investigated the M•–σ relation using a ies of nearby galaxies show that the reddening obtained from NIR sample of 93 late-type galaxies with a Seyfert 1 nucleus. They found

MNRAS 446, 2823–2836 (2015) 2832 R. A. Riffel et al. no difference in the slope for subsamples of barred and unbarred logarithmic difference of −0.29 ± 0.12, which must be considered galaxies, but they found a small offset in the relation between low- as a systematic error in the SMBH mass when using σ CO for spiral and high-inclination disc galaxies, with the latter having a larger σ galaxies. However, this uncertainty is dominated by scatter of the value for a given black hole mass. For a review of calibrations of relation and the conversion from σ CO to σ opt may introduce an even the M•–σ relation, see Kormendy & Ho (2013). larger uncertainty in the derived M•. Actually, Kormendy, Bender & Cornell (2011) show that the Although the ‘σ -discrepancy’ has already been discussed for physically relevant parameter in black hole correlations with host ULIRGs and merger remnants (e.g. Rothberg & Fischer 2010; Roth- galaxy type is not early-type versus late-type objects, but rather berg et al. 2013) and no discrepancy was found for early-type galax- classical bulges versus pseudo-bulges. The latter is defined as the ies (e.g. Silge & Gebhardt 2003;Vanderbekeetal.2011; Kang et al. buildup of dense central components that look like classical merger- 2013), this is the first time that this comparison is done for a sample built bulges but that were in fact formed slowly by disc out of disc of mostly late-type galaxies. material (Kormendy & Kennicutt 2004). Since the aim of the present paper is to evaluate the impact of the use of CO-based measurements of the stellar velocity disper- ACKNOWLEDGEMENTS sion on the derived mass of the SMBH, and not to calibrate the We thank an anonymous referee for useful suggestions which helped σ M•– relationship, we use the same calibration for all objects of to improve the paper and S. Rembold for help with the bootstrap our sample, given by Kormendy & Ho (2013)as   technique. This work is based on observations obtained at the Gem- Downloaded from M• ini Observatory, which is operated by the Association of Universities log =−(0.500 ± 0.049) 109 M for Research in Astronomy, Inc., under a cooperative agreement   with the NSF on behalf of the Gemini partnership: the National σ + (4.420 ± 0.295)log . (4) Science Foundation (United States), the Science and Technology 200 km s−1 Facilities Council (United Kingdom), the National Research Coun- http://mnras.oxfordjournals.org/ We estimated the mass of the SMBH for all spiral galaxies of cil (Canada), CONICYT (Chile), the Australian Research Council (Australia), Ministerio´ da Ciencia,ˆ Tecnologia e Inovac¸ao˜ (Brazil), our sample using σ CaT and σ CO in the equation above. The mean logarithmic difference is −0.29 ± 0.12, which may be taken as a and Ministerio de Ciencia, Tecnologia e Innovacion´ Productiva systematic error in the M•–σ relation when using CO-based esti- (Argentina). RAR acknowledges support from FAPERGS (project mates of the stellar velocity dispersion. no. 12/1209-6) and CNPq (project no. 470090/2013-8). LCH acknowledges support from the Kavli Foundation, Peking Uni- versity, and grant no. XDB09030102 (Emergence of Cosmolog- 7 CONCLUSIONS ical Structures) from the Strategic Priority Research Program of We have used 0.85–2.5 µm spectroscopy of a sample of 48 galaxies the Chinese Academy of Sciences. ARA acknowledges CNPq

(35 spirals, 7 lenticulars, and 6 ellipticals) obtained with the GNIRS for partial support to this work through grant 307403/2012-2. at Universidad de Chile on May 20, 2015 on Gemini North telescope to measure the stellar kinematics by LM thanks CNPq through grant 305291/2012-2. LC acknowl- fitting the K-band CO absorption band heads and the CaT at 8550 Å. edges support from the Special Visiting Researcher Fellowship This work is aimed at determining whether the difference in σ CO (PVE 313945/2013-6) under the Brazilian Scientific Mobility Pro- and σ CaT (the ‘σ -discrepancy’) reported for ULIRGs and merger gram ‘Cienciasˆ sem Fronteiras’. RR acknowledges funding from remnants persists in the hitherto unexplored regime of late-type FAPERGs (ARD 11/1758-5) and CNPq (PeP 304796/2011-5). CRA galaxies. Our main conclusions are as follows. is supported by a Marie Curie Intra European Fellowship within the 7th European Community Framework Programme (PIEF-GA- µ (i) The velocity dispersion obtained from the 2.29 mCOband 2012-327934) and by the Spanish Ministry of Science and Innova- heads is slightly smaller than the one from fitting the CaT, with tion (MICINN) through project PN AYA2010-21887-C04.04. −1 an average difference of σ CO − σ CaT =−19 ± 6kms for the complete sample (all morphological types). (ii) We compiled the available σ values from the literature and REFERENCES found an almost one-to-one relation between optical (CaT, Mgb, etc.) and CO-based estimates for early-type galaxies. For spi- Auld R. et al., 2013, MNRAS, 428, 1880 Barth A. J., Ho L. C., Sargent W. L. W., 2002a, ApJ, 124, 2607 ral galaxies, the discrepancy is higher, but still much lower than σ = Barth A. J., Ho L. C., Sargent W. L. W., 2002b, ApJ, 566, L13 for merger remnants. The best fit for spiral galaxies is opt Beers T. C., Flynn K., Gebhardt K., 1990, AJ, 100, 32 ± + ± σ (46.0 18.1) (0.85 0.12) CO, but more observations cov- Bellovary J., Holley-Bockelmann K., Gultekin¨ K., Christensen C., −1 −1 ering the σ ranges σ<100 km s and σ>200 km s are needed Governato F., Brooks A., Loebman S., Munshi F., 2014, MNRAS, 445, to properly calibrate this relation. 2667 (iii) The fractional σ difference correlates with the IR luminos- Bender R., Burstein D., Faber S. M., 1993, ApJ, 411, 153 ity, which may suggest that the σ -discrepancy is related to the Bernardi M. et al., 2003, AJ, 125, 1866 presence of warm dust. In this scenario, the CO absorption band Bower R. G., Benson A. J., Malbon R., Helly J. C., Frenk C. S., Baugh C. heads would be dominated by young stars located in the disc of M., Cole S., Lacey C. G., 2006, MNRAS, 370, 645 the galaxy and thus result in smaller σ values, while the optical Bruzual G., Charlot S., 2003, MNRAS, 344, 1000 Cappellari M., Emsellem E., 2004, PASP, 116, 138 estimates are less sensitive to variations in the stellar population. Cappellari M. et al., 2007, MNRAS, 379, 418 However, the detailed spectral synthesis that would be needed to test Cenarro A. J., Cardiel N., Gorgas J., Peletier R. F., Vazdekis A., Prada F., this interpretation requires high spectral resolution Simple Stellar 2001, MNRAS, 326, 959 Populations (SSP) models, which are not yet available. Ciesla L. et al., 2012, A&A, 543, 161 (iv) We investigated the impact of the σ -discrepancy on the mass Cox T. J., Dutta S. N., Di Matteo T., Hernquist L., Hopkins P. F., Robertson of the SMBH obtained via the M•–σ relation and found a mean B., Springel V., 2006, ApJ, 650, 791

MNRAS 446, 2823–2836 (2015) The σ -discrepancy in spiral galaxies 2833

Cushing M. C., Rayner J. T., Vacca W. D., 2005, ApJ, 623, 1115 Martins L. P., Rodr´ıguez-Ardila A., Diniz S., Gruenwald R., de Souza R., De Looze I. et al., 2012, MNRAS, 423, 2359 2013a, MNRAS, 431, 1823 Di Matteo T., Springel V., Hernquist L., 2005, Nature, 433, 604 Moorwood A. F. M., Oliva E., 1988, A&A, 203, 278 Djorgovski S., Davis M., 1987, ApJ, 313, 59 Naab T. et al., 2013, MNRAS, 444, 3357 Draine B. T., 2003, ARA&A, 41, 241 Nemmen R., Bower R., Babul A., Storchi-Bergmann T., 2007, MNRAS, Dressler A., 1984a, ApJ, 281, 512 377, 1652 Dressler A., Lynden-Bell D., Burstein D., Davies R. L., Faber S. M., Origlia L., Moorwood A. F. M., Oliva E., 1993, A&A, 280, 5360 Terlevich R., Wegner G., 1987, ApJ, 313, 42 Rayner J. T., Cushing M. C., Vacca W. D., 2009, ApJS, 185, 289 Emsellem E., Greusard D., Combes F., Friedli D., Leon S., Pecontal´ E., Riffel R. A., 2010, Ap&SS, 327, 239 Wozniak H., 2001, A&A, 368, 52 Riffel R., Rodr´ıguez-Ardila A., Pastoriza M. G., 2006, A&A, 457, 61 Emsellem E. et al., 2004, MNRAS, 352, 721 Riffel R., Pastoriza M. G., Rodr´ıguez-Ardila A., Maraston C., 2007, ApJ, Falcon-Barroso´ J. Peletier R. F., Balcells M., 2002, MNRAS, 335, 741 659, L103 Ferrarese L., Merrit D., 2000, ApJ, 547, 140 Riffel R. A., Storchi-Bergmann T., Winge C., McGregor P. J., Beck T., Gebhardt K. et al., 2000, ApJ, 539, 13 Schmitt H., 2008, MNRAS, 385, 1129 Gebhardt K. et al., 2003, ApJ, 583, 92 Riffel R. A., Storchi-Bergmann T., Riffel R., Pastoriza M. G., 2010, ApJ, Gultekin¨ K., Cackett E. M., Miller J. M., Di Matteo T., Markoff S., Richstone 713, 469 D. O., 2009, ApJ, 706, 404 Riffel R., Riffel R. A., Ferrari F., Storchi-Bergmann T., 2011, MNRAS, 416, Heisler C. A., De Robertis M. M., 1999, AJ, 118, 2038 493

Hildebrand R. H., 1983, QJRAS, 24, 267 Rothberg B., Fischer J., 2010, ApJ, 712, 318 Downloaded from Ho L. C., Filippenko A. V., Sargent W. L., 1995, ApJS, 98, 477 Rothberg B., Fischer J., Rodrigues M., Sanders D. B., 2013, ApJ, 767, 72 Ho L. C., Filippenko A. V., Sargent W. L., 1997, ApJS, 112, 31 Sanchez-Bl´ azquez´ P. et al., 2006, MNRAS, 371, 703 Ho L. C., Greene J. E., Filippenko A. V., Sargent W. L., 2009, ApJS, 183, 1 Sauvage M. et al., 2010, A&A, 518, 64 Ivanov V. D., Rieke G. H., Groppi C. E., Alonso-Herrero A., Rieke M. J., Silge J. D., Gebhardt K., 2003, ApJ, 125, 2809 Engelbracht C. W., 2000, ApJ, 545, 190 Silva D. R., Kuntschner H., Lyubenova M., 2008, ApJ, 674, 194

Kang W.-R., Woo J-H., Schulze J., Riechers D. A., Kim S. C., Park D., Springel V., Di Matteo T., Hernquist L., 2005, ApJ, 620, 79 http://mnras.oxfordjournals.org/ Smolcic V., 2013, ApJ, 767, 26 Storchi-Bergmann T., Riffel R. A., Riffel R., Diniz M. R., Borges Vale T., Kennicutt R. C., 1998, ARA&A, 36, 189 McGregor P. J., 2012, ApJ, 755, 87 Kennicutt R. C., Evans N. J., 2012, ARA&A, 50, 531 Terlevich R., Davies R. L., Faber S. M., Burstein D., 1981, MNRAS, 196, Kormendy J. Bender R., Cornell M. E., 2011, Nature, 469, 374 381 Kormendy J., Ho L. C., 2013, ARA&A, 51, 511 Thuan T. X., Sauvage M., 1992, A&AS, 92, 749 Kormendy J., Kennicutt R., 2004, ARA&A, 42, 603 Valluri M., Merritt D., Emsellem E., 2004, ApJ, 602, 66 Kotilainen J. K., Hyvonen¨ T., Reunanen J., Ivanov V. D., 2012, MNRAS, van der Marel R. P., Franx M., 1993, ApJ, 407, 525 425, 1057 Vanderbeke J., Baes M., Romanowsky A. J., Schimidtobreick L., 2011, Kuntschner H., 2000, MNRAS, 315, 184 MNRAS, 412, 2017 Maraston C., 2005, MNRAS, 362, 799 Weingartner J. C., Draine B. T., 2001, ApJ, 548, 296 at Universidad de Chile on May 20, 2015 Marleau F. R. et al., 2006, ApJ, 646, 929 Winge C., Riffel R. A., Storchi-Bergmann T., 2009, ApJS, 185, 186 Marquez´ I. Masegosa J., Durret F., Gonzalez´ Delgado R. M., Moles M., Xiao T., Barth A. J., Greene J. E., Ho L. C., Bentz M. C., Ludwig R. R., Maza J., Perez´ E., Roth M., 2003, A&A, 409, 459 Jiang Y., 2011, ApJ, 739, 28

MNRAS 446, 2823–2836 (2015) 2834 R. A. Riffel et al.

APPENDIX A: FITS OF THE SPECTRA Figs A1 and A2 show the resulting fit of the galaxy spectra for the CO and Ca triplet spectral regions, respectively. Downloaded from

Figure A1. Sample fits of the CO absorptions band heads. The observed spectra are shown as dotted lines and the best-fitting model as continuous lines. In the bottom panels, the residuals of the fits (observed model) are shown as dotted lines, where the dashed lines show the 1σ level of the continuum. Fluxes are shown in units of 10−15 erg s−1 cm−2 Å −1 and the residuals in units of 10−17 erg s−1 cm−2 Å −1.Theadev parameter shown at each panel gives the percentage | |/

mean OM O deviation over all fitted pixels, where O is the observed spectrum and M is the model. The fits for all spectra are available online only. http://mnras.oxfordjournals.org/ at Universidad de Chile on May 20, 2015

Figure A2. Same as Fig. A1 for the Ca triplet.

MNRAS 446, 2823–2836 (2015) The σ -discrepancy in spiral galaxies 2835

APPENDIX B: STELLAR TEMPLATES USED TO FIT THE STELLAR KINEMATICS Table B1 shows the weights of each star (as well their spectral types) to the fit of the galaxy spectra for the CaT and CO spectral regions.

Table B1. Sample of the stellar templates used for each galaxy to derive the stellar kinematics. Column 1: galaxy name; columns 2–4: spectral and luminosity class, name and percentage contribution of the star to the fit of the galaxy spectrum for the Ca triplet region. Templates are from Cenarro et al. (2001) and for some cluster stars, the authors list the positions of the stars in the HR diagram (SGB: subgiant branch; GB: giant branch and HB: horizontal branch). Columns 5–7: same as columns 2–4 for the CO spectral region. Templates are from Winge et al. (2009). The full table is available online only.

CaT region CO region Galaxy Spectral type Star Weight (per cent) Spectral type Star Weight (per cent)

NGC 1358 HB M71-1-34 44 M2III HD30354 57 HB M71-C 9 M2 BD+59 274 7 Downloaded from K4III HD149161 6 K8V HD113538 33 M5III HD172816 24 M0III HD2490 1 M7III HD114961 15 NGC 1961 GB NGC 188-I-57 27 M2III HD30354 36 K0V HD149661 9 K8V HD113538 24

K5III HD139669 39 M0III HD2490 39 http://mnras.oxfordjournals.org/ M5.5III HD94705 19 M6V BD+19-5116-B 2 M7III HD114961 1 NGC 2273 HB M5-II-53 8 M2III HD30354 18 K4II HD130705 39 K3Iab HD339034 1 K5III HD139669 1 K8V HD113538 1 M4III HD17491 20 M0III HD2490 19 M5III HD172816 8 K0IV HD34642 4 M5.5III HD94705 1 G5II HD36079 33

M5III HD175865 7 K7III HD63425B 3 at Universidad de Chile on May 20, 2015 SGB M67-F-115 10 G8V HD64606 11 SGB M67-F-125 5 G3V HD6461 6 NGC 2655 F3III HD115604 6 M2III HD30354 7 GB M67-F-108 7 K3Iab HD339034 1 HB M5-II-53 8 K8V HD113538 25 K5III HD139669 31 M0III HD2490 40 M6III HD18191 22 K0IV HD34642 1 SGB M67-F-125 24 K7III HD63425B 23 NGC 2768 GB NGC 188-I-85 8 M2III HD30354 23 HB M71-1-41 16 K3Iab HD339034 1 HB M92-XII-24 4 K8V HD113538 19 K3III HD102328 37 K1II HD198700 3 M6III HD18191 24 M0III HD2490 49 SGB M67-F-125 7 K0IV HD34642 2 NGC 2832 F0 BD-01-2582 1 1 HB M71-C 41 M2III HD30354 30 HB M92-XII-24 2 K8V HD113538 17 M5.5III HD94705 20 M0III HD2490 51 M6V BD+19-5116-B 16 M7III HD207076 6 SGB M67-F-125 10 NGC 3031 GB M67-F-231 1 M3III HD27796 5 GB M92-XII-8 2 M2III HD30354 14 GB NGC 188-II-122 6 K8V HD113538 21 HB M5-II-76 1 M0III HD2490 10 HB M92-I-13 3 K0IV HD34642 2 K5III HD139669 5 G5II HD36079 8 M5III HD172816 25 K7III HD63425B 25 M7III HD114961 18 G3V HD6461 10 SGB M67-F-115 7 K4III HD9138 1 SGB M67-IV-68 29

MNRAS 446, 2823–2836 (2015) 2836 R. A. Riffel et al.

8Gemini Observatory, Southern Operations Center, Casilla 603, La Serena, SUPPORTING INFORMATION Chile Additional Supporting Information may be found in the online ver- 9Astrophysics Department, Center for Astrobiology (CSIC-INTA), Torrejon sion of this article: de Ardoz, E-28850 Madrid, Spain 10Instituto de Fisica de Cantabria, CSIC-UC, E-39005 Santander, Spain Appendix A: Fits of the spectra. 11Department of Physics, University of the Pacific, 3601 Pacific Avenue, Appendix B: Stellar templates used to fit the stellar kinematics. Stockton, CA 95211, USA (http://mnras.oxfordjournals.org/lookup/suppl/doi:10.1093/mnras/ 12Instituto de Astrof´ısica de Canarias, Calle V´ıa Lactea,´ s/n, E-38205 La stu2256/-/DC1). Laguna, Tenerife, Spain 13Departamento de Astrof´ısica, Universidad de La Laguna, E-38205 La Please note: Oxford University Press are not responsible for the Laguna, Tenerife, Spain content or functionality of any supporting materials supplied by 14Centro de Radioastronoma y Astrofsica (UNAM), Apartado Postal 3-72 the authors. Any queries (other than missing material) should be (Xangari), 58089 Morelia, Mexico directed to the corresponding author for the paper. 15Departamento de Astronom´ıa, Universidad de Chile, Casilla 36-D, San- tiago, Chile 16Department of Physics and Astronomy, Macquarie University, NSW 2109, 1Departamento de F´ısica/CCNE, Universidade Federal de Santa Maria, Australia 97105-900, Santa Maria, RS, Brazil 17Astrophysics Research Institute, Liverpool John Moores University, IC2, 2Kavli Institute for Astronomy and Astrophysics, Peking University, Beijing Liverpool Science Park 146 Brownlow Hill, Liverpool L3 5RF, UK Downloaded from 100871, China 18Department of Physics & Astronomy, University of Victoria, PO Box 1700, 3Department of Astronomy, School of Physics, Peking University, Beijing STN CSC Victoria, BC V8W 2Y2, Canada 100871, China 19Department of Physics and Space Sciences, Florida Institute of Technol- 4Gemini Observatory, Northern Operations Center, 670 N. Aohoku Place, ogy, 150 West University Boulevard, Melbourne, FL 32901, USA Hilo, HI 96720, USA 5Laboratorio´ Nacional de Astrof´ısica/MCT, Rua dos Estados Unidos 154, Itajuba,´ MG, Brazil http://mnras.oxfordjournals.org/ 6NAT – Universidade Cruzeiro do Sul, Rua Galvao˜ Bueno, 868, Sao˜ Paulo 01506-000, SP, Brazil 7Instituto de F´ısica, Universidade Federal do Rio Grande do Sul, CP 15051, Porto Alegre 91501-970, RS, Brazil This paper has been typeset from a TEX/LATEX file prepared by the author. at Universidad de Chile on May 20, 2015

MNRAS 446, 2823–2836 (2015)